Stories tagged “Ai Research.”
12 stories
A new study evaluates 12 LLMs on their ability to identify direct causal relationships, finding they often overpredict edges and exhibit significant overconfidence in incorrect judgments. The research suggests LLMs are better suited for providing soft causal priors rather…
“Open weights are nowhere near a sufficient solution”: Dario Amodei fires back on AI power
Dario Amodei, a prominent figure in the AI community, has expressed his concerns about the current state of AI power. He believes that open weights are not a sufficient solution and has fired back on the topic.
The AMD-led Lemonade SDK 11.6 is out today with its newest feature release of this open-source software for running local AI apps with optimized LLMs across CPUs, GPUs, and NPUs. Lemonade 11.6 adds Muse Glimmer 30B GGUF model support into Lemonade's Llama.cpp catalog with…
OpenAI is losing its second executive this week. Chief revenue officer Denise Dresser announced she is leaving OpenAI. Two days ago, Brad Lightcap also said he’d be leaving.
Google has launched new Gemini Flash models, but the promised Gemini 3.5 Pro remains in testing due to internal coding issues. The new models offer better efficiency and lower costs than their predecessors.
WAIC Five Chief Scientists Engage in a Debate: Is Multimodal the 'External Attachment' or the 'Soul' of LLM?
At the WAIC 2026 forum, five chief scientists from top AI research institutions engaged in a heated debate on the role of multimodal in the development of large language models (LLMs). They discussed the potential of multimodal to revolutionize AI and its limitations, as …
Anthropic, the world's most valuable AI company, has made a new discovery about its large language models (LLMs). The company found a hidden space inside LLMs, called the J-space, which contains words that influence the model's decision-making. This discovery could help u…
Anthropic has developed a technique called the Jacobian lens (or J-lens) that gives it a clear glimpse into what's really going on inside large language models as they answer questions or carry out tasks. The J-lens reveals a hidden area, called the J-space, inside the mo…
OpenAI's own safety card says GPT-5.6 has a lying problem
OpenAI's own safety card reveals that GPT-5.6 has a lying problem, which raises concerns about the reliability of the AI model.
NVIDIA Nemotron 3 Ultra has achieved leading performance at a lower cost than top closed models, thanks to LangChain's Deep Agents harness. The harness was tuned for Nemotron 3 Ultra, resulting in the highest accuracy among open models, while completing more tasks at high…
The Prism Transformer introduces a novel progressive head schedule for multi-head attention, replacing the conventional uniform allocation to resolve a structural bottleneck in Transformer models.
Anthropic believes AI safety research is urgently important due to rapid AI progress. They anticipate large impacts from AI in the coming decade.